Your organization may be human-readable. Is it agent-ready?

On 1 October 2026, Kyndryl opened its AI Innovation Lab in Bertrange, Luxembourg — its first in the European Union. BIL was announced as its first client and partner.

The Lab starts with real business problems and works across people, processes and technology — including the workflows and operating models required to bring agentic AI into real organizational environments.

At first glance, this is another signal of AI accelerating in Luxembourg.

But something more interesting may be happening underneath.

Until now, much of enterprise AI has helped us do things: write, analyze, search, summarize, produce.

Agentic AI introduces a different possibility: to act.

And there is a profound difference between helping someone perform a task and delegating the capacity to act.

When AI moves from answering to acting, the cost of what remains implicit changes.

Because companies do not run on processes alone.

They also run on everything the people inside them have learned to understand.

Someone knows that this particular customer is an exception.

Someone else knows that above a certain threshold, a decision really needs to go back to the CEO — even if no rule says so explicitly.

A colleague can spot the authoritative file among three almost identical versions.

Another recognizes that an apparently ordinary request is urgent because they understand the context around it.

People ask. They observe. They interpret. They remember. They learn who really decides what.

And very often, people who know an organization compensate for what the organization has never needed to represent completely.

That is not necessarily a failure of organization. It is one of the reasons companies have been able to operate this way for years.

Then a new actor enters the workflow.

An agent.

And an apparently technical question becomes an organizational one:

What does it need to know in order to act?

The obvious answer would be: everything that used to be implicit must now become explicit.

Except that this is not quite true.

Agentic systems can acquire context dynamically. AI can search across sources, retrieve information and, in some cases, even help extract and formalize knowledge that was previously tacit.

So the challenge is probably not to turn the entire company into a procedure manual.

An organization does not need to make everything explicit. It needs to make the right things intelligible enough for someone — or something — to act.

Which raises a more useful set of questions.

Who does what — and who actually owns the outcome?

Who can decide what? With what authority?

Which information should be trusted when several versions coexist?

Does a piece of information mean the same thing to everyone — or does its meaning still depend on who reads it?

What is the rule — and what happens when reality becomes the exception?

What is someone, or something, actually allowed to do? Where does that permission stop?

When should a case be escalated? To whom? And how does work move from one actor to the next?

How do we know what was decided, executed and controlled — and who remains accountable?

And finally: what can an experienced person still recognize, interpret or arbitrate without being able to explain it easily?

These questions do not describe a single process.

They reveal multiple dimensions of how an organization actually works: roles and ownership; decision rights and authority; information and sources of truth; semantic context; rules, exceptions and arbitration; permissions and boundaries; escalation and handovers; control, traceability and accountability; tacit knowledge and contextual judgment.

And this is where the story stops being only about AI.

Replace the agent with a new employee.

Or with someone to whom the founder finally wants to delegate part of their role.

Replace it with an ERP implementation.

A temporary replacement.

A growing team.

Or a successor trying to understand a business without having lived through the ten years that shaped it.

How many of those questions really disappear?

What if the organizational problem revealed by agentic AI existed long before AI?

At DEAMERA SME Transformation, we use a term for this multidimensional reality: Organizational Intelligibility.

It is not another name for an Operating Model.

It is not a BPMN diagram.

It is not a RACI matrix.

And it certainly does not mean documenting everything a company does.

An Operating Model, a BPMN diagram, a RACI matrix, a process map, an ERP or a body of documentation can each represent or structure parts of this reality.

But none of them, on its own, answers the more fundamental question:

Is the organization intelligible enough for the actor entrusted with an action to understand what to do, which information to rely on, what authority it has, where its boundaries lie — and what should happen when reality no longer fits the standard case?

That is what we mean here by Organizational Intelligibility: an organization’s capacity to make the context, roles, authority, decisions, rules, exceptions, permissions, handovers and accountability required for action sufficiently accessible, understandable and governable.

The important word may be sufficiently.

Because a living organization will always contain context, interpretation and some degree of implicit knowledge.

And because AI itself may help make some of that knowledge more accessible.

The question, then, is not how to make everything explicit.

It is:

What needs to become intelligible? To whom — or to what? For which action? And at what level of risk?

This is why Agent Readiness may prove far more interesting than another technology-readiness checklist.

Our hypothesis is that it could become a particularly powerful contemporary stress-test of a much older organizational capability.

A company can function because its people continuously fill in the blanks.

It can function very well that way.

But when it starts delegating more — to a new employee, a system, a successor or, now, an AI agent — those blanks change in nature.

They become visible.

And sometimes costly.

AI may not have created the problem at all.

It may simply have changed the cost and risk of what the organization could previously afford to leave implicit.

So perhaps the first question to ask before wondering what an AI agent could do inside your company is not:

“What can AI do for us?”

It is:

What must your organization make accessible, reliable and governable when the next actor in the workflow is no longer only human?

AI may be the doorway to your company's next evolution. But before you cross it, your organization may need to become intelligible to itself.